Home-based care information management method and system based on Internet of Things

Abstract

The invention relates to a home-based care information management method and system based on the Internet of Things. The method comprises the steps of obtaining home-based care original data corresponding to the home-based elderly in response to an information access request for the home-based elderly; performing data exception suppression processing on the home-based care original data to obtain home-based care credible data; performing event semantic analysis on the home-based care state according to the home-based care credible data to obtain a care semantic event set and a personal care knowledge graph; according to the endowment semantic event set and the personal endowment knowledge graph, edge-side reasoning is carried out on the home-based endowment credible data, and an elderly disclosure conclusion set is obtained; according to the information access request, carrying out situation contract judgment on an information disclosure strategy to obtain information disclosure analysis data; and according to the information disclosure analysis data, carrying out verifiable minimum disclosure analysis on the elderly disclosure conclusion set to obtain home-based care output information and a home-based care verification proof. According to the method, privacy leakage can be remarkably reduced.

Description

Technical Field

[0001] This application relates to the field of intelligent management technology, and in particular to a method and system for home-based elderly care information management based on the Internet of Things. Background Technology

[0002] Traditional home-based elderly care information management methods typically rely on community / elderly care service platforms and home terminals (such as mobile apps, hotline operators, or community workstations) to centrally collect, store, and transfer information. Basic profiles (identity information, living conditions, emergency contacts, chronic diseases and medications, service preferences, etc.) are entered by the elderly person, their family, community grid workers, or caregivers. The platform establishes unified electronic profiles and distributes them to family doctors, elderly care institutions, and home service providers according to their permissions. In daily operation, requests for meal assistance, cleaning assistance, bathing assistance, medical accompaniment, and rehabilitation are handled through appointment work orders, forming a closed-loop record of "application—dispatch—home visit—follow-up visit—evaluation." Health and service data are updated through telephone follow-ups or regular visits. In case of emergency, the platform can pre-set emergency contacts and service resources in the profile. Upon triggering an alarm, operators / community personnel will coordinate home visits or transfer to emergency services, enabling basic information retrieval, statistical reports, and service process documentation. However, traditional technologies rely heavily on manual data entry and cross-entity data sharing, which can easily lead to privacy leaks and unauthorized access. Summary of the Invention

[0003] Therefore, it is necessary to provide an IoT-based home-based elderly care information management method and computer equipment that can significantly reduce the risks of privacy leaks and unauthorized access, addressing the aforementioned technical issues.

[0004] Firstly, this application provides a method for managing home-based elderly care information based on the Internet of Things, including: In response to an information access request for an elderly person living at home, obtain the original home-based elderly care data corresponding to the elderly person living at home; Data anomaly suppression processing is performed on the original home-based elderly care data to obtain reliable home-based elderly care data; Based on the reliable data on home-based elderly care, the home-based elderly care status is analyzed using event semantics to obtain a set of elderly care semantic events and a personal elderly care knowledge graph. Based on the set of semantic events related to elderly care and the personal knowledge graph of elderly care, edge-side reasoning is performed on the trusted data of home-based elderly care to obtain a set of conclusions disclosed by the elderly. Based on the information access request, a situational contract adjudication is performed on the information disclosure strategy of the elderly living at home to obtain information disclosure analysis data. Based on the information disclosure analysis data, a verifiable minimum disclosure analysis is performed on the set of elderly disclosure conclusions to obtain home-based elderly care output information and home-based elderly care verification certificates.

[0005] Secondly, this application also provides a home-based elderly care information management system based on the Internet of Things, including: computer equipment and triggering terminal; The triggering terminal is used to respond to an information access request for an elderly person living at home and obtain the original home-based elderly care data corresponding to the elderly person; the original home-based elderly care data is transmitted to the computer device via a network; The computer equipment is used to perform data anomaly suppression processing on the original home-based elderly care data to obtain reliable home-based elderly care data. The computer device is used to perform event semantic analysis on the home-based elderly care status of the elderly based on the trusted home-based elderly care data, and obtain a set of elderly care semantic events and a personal elderly care knowledge graph. The computer device is used to perform edge-side reasoning on the trusted home-based elderly care data based on the set of elderly care semantic events and the personal elderly care knowledge graph, to obtain a set of conclusions disclosed by the elderly. The computer device is used to make a situational contract ruling on the information disclosure strategy of the elderly living at home based on the information access request, and obtain information disclosure analysis data. The computer device is used to perform verifiable minimum disclosure analysis on the set of elderly disclosure conclusions based on the information disclosure analysis data, and to obtain home-based elderly care output information and home-based elderly care verification certificates.

[0006] The aforementioned IoT-based home-based elderly care information management method and system improves data reliability by suppressing anomalies in raw home-based elderly care data after receiving information access requests. Based on trusted data, it completes event semanticization and constructs a personal elderly care knowledge graph, transforming the home-based elderly care status from "raw collected data" into computable, associative, and interpretable semantic events and structured knowledge. This allows for reasoning at the edge using semantic events and knowledge graphs to obtain a high-confidence set of elderly disclosure conclusions. Furthermore, by introducing contextual contract adjudication oriented towards identity, purpose, and situation, it generates information disclosure analysis data, achieving granular disclosure. Dynamic constraints on scope, timeliness, and retrieval conditions transform the output content from "full data sharing" to "on-demand, minimal" disclosure of conclusions / summaries / fragments. Simultaneously, verifiable minimal disclosure analysis generates home-based elderly care output information and its verification proof, enabling external requesters to verify the source consistency and compliance of the output information without accessing the original reproducible data. This significantly reduces the risk of privacy leaks and unauthorized access, improves the credibility, traceability, and cross-entity collaboration efficiency of home-based elderly care information sharing, reduces cloud or platform-side storage and computing pressure, and enhances the system's continuous service capabilities in weak network or offline scenarios. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 This is an application environment diagram of an IoT-based home-based elderly care information management method in one embodiment; Figure 2 This is a flowchart illustrating an IoT-based home-based elderly care information management method in one embodiment. Figure 3 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0009] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0010] This application provides an Internet of Things-based home-based elderly care information management method, which can be applied to, for example... Figure 1 In the application environment shown, the trigger terminal 102 communicates with the computer device 104 via a network. A data storage system can store the data that the computer device 104 needs to process. The data storage system can be integrated onto the computer device 104, or it can be located in the cloud or on other networked computer devices. The trigger terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. Portable wearable devices can include smartwatches, smart bracelets, head-mounted devices, etc. The computer device 104 can be implemented using a standalone computer device or a cluster of multiple computer devices.

[0011] In one exemplary embodiment, such as Figure 2 As shown, a method for managing home-based elderly care information based on the Internet of Things is provided, which can be applied to... Figure 1 The following steps are used as an example of computer equipment, including steps 202 to 212. Wherein:

[0012] Step 202: In response to the information access request for elderly people living at home, obtain the original home-based elderly care data corresponding to the elderly people living at home.

[0013] Step 204: Perform data anomaly suppression processing on the original home-based elderly care data to obtain reliable home-based elderly care data.

[0014] Step 206: Based on reliable data on home-based elderly care, perform event semantic analysis on the home-based elderly care status to obtain a set of elderly care semantic events and a personal elderly care knowledge graph.

[0015] Step 208: Based on the set of semantic events related to elderly care and the personal knowledge graph of elderly care, perform edge-side reasoning on the credible data of home-based elderly care to obtain the set of conclusions disclosed by the elderly.

[0016] Step 210: Based on the information access request, conduct a situational contract adjudication on the information disclosure strategy for elderly people living at home, and obtain information disclosure analysis data.

[0017] Step 212: Based on the information disclosure analysis data, perform verifiable minimum disclosure analysis on the set of disclosure conclusions of the elderly to obtain home-based elderly care output information and home-based elderly care verification certificates.

[0018] Among them, an information access request is a request message initiated by an external requester to obtain elderly care information of elderly people living at home, and it carries at least identity credentials, access purpose, time range and context parameters.

[0019] Among them, the raw data for home-based elderly care is raw time-series data, event data, or multimedia data directly collected by IoT sensing devices in the home environment without being cleaned or semantically converted.

[0020] Among them, data anomaly suppression processing is a data cleaning process that identifies, downweights, corrects, or removes anomalies such as noise, outliers, drift, missing data, and duplicate reporting in the original data.

[0021] Among them, trusted data for home-based elderly care is a standardized data set that has undergone anomaly suppression and is accompanied by trustworthiness labels / scoring, and can be used for subsequent analysis and reasoning.

[0022] Among them, the home-based elderly care status is a description of the comprehensive status of elderly people living at home, including their behavior, physiology, medication, and environment at present or for a period of time.

[0023] Among them, semantic event analysis is the analysis and processing that maps credible data into elderly care scenario events with clear meaning and extracts event attributes (time, location, object, confidence level, etc.).

[0024] Among them, the elderly care semantic event set is a structured list or event stream obtained from event semantic analysis, used to describe elderly care activities and anomalies.

[0025] Among them, the personal elderly care knowledge graph is a structured association graph constructed with "time-place-activity-physiology-medication-environment-risk-intervention" as entities and relationships, used to express the elderly person's elderly care knowledge and evidence chain.

[0026] Edge-side inference is the process of performing inferential computations on home edge devices based on semantic events, knowledge graphs, and trusted data to generate risk conclusions or intervention recommendations.

[0027] Among them, the elderly disclosure conclusion set is a set of conclusions output by the marginal side reasoning that can be disclosed to the outside world. Each conclusion usually includes a conclusion value, confidence level and evidence index.

[0028] Among them, the information disclosure strategy is a set of strategy rules used to specify the granularity, scope, timeliness, and retrieval conditions of information that can be disclosed under different identities, purposes, and situations.

[0029] Among them, the situational contract adjudication is a process of dynamically compiling and deciding on the information disclosure strategy based on the identity and purpose of the requester, combined with the situational risk and urgency, to determine the boundaries of this disclosure.

[0030] Among them, the information disclosure analysis data is the output data of the situational contract ruling, which includes at least the disclosure level, disclosure scope, validity period, recall trigger conditions and proof requirements.

[0031] Verifiable minimum disclosure analysis is a process that selectively outputs the minimum necessary information and generates verifiable materials to prove compliance and consistency, while meeting the data constraints of disclosure analysis.

[0032] Among them, the information output by home-based elderly care is the external output content generated in accordance with the principle of minimum disclosure, which is usually the conclusion, statistical summary and / or explanatory fragments rather than the content that can restore the original data.

[0033] Among them, the home-based elderly care verification certificate is verifiable evidence data used to prove that the home-based elderly care output information is generated by the edge side according to the situational contract constraints and has not been tampered with, and is consistent with the commitment.

[0034] Specifically, upon receiving an external request for information access from an elderly person living at home, the triggering terminal converts the request content into a request message, records it, and transmits it over the network to a computer device. The computer device then parses the request message to extract the elderly person's identifier, the requester's identity credentials, the purpose of access, the time range, and contextual parameters associated with the request. At the data access layer on the home edge, the corresponding triggering terminal responds to the computer device's extraction request. Based on the elderly person's identifier and the time range, it locates the corresponding original home-based elderly care data from the data cache of connected IoT sensing devices, the edge-side streaming event bus, and / or the local personal data vault. It then uniformly encapsulates the original data from different sources with data source tags and timestamps, forming original home-based elderly care data that corresponds one-to-one with this information access request.

[0035] Computer equipment performs noise and anomaly identification on different data types in the raw data of home-based elderly care. Specifically, for continuous physiological / environmental time series, threshold + sliding window statistics or robust filtering are used to detect peaks, drifts and abrupt changes. For discrete events, frequency / time series consistency rules are used to identify duplicate reports and false triggers. Offline, missing and outlier segments are imputed, downweighted or removed. At the same time, a credibility score is attached to each data point (e.g., based on equipment health, sampling completeness, anomaly ratio). The final output includes numerical correction results and credibility labels for credible home-based elderly care data.

[0036] Computer devices map trusted home-based elderly care data as multi-source data to a unified elderly care semantic event space. For example, they can identify "leaving bed / going to the toilet / returning to bed" based on body movement, mattress pressure, and door sensor combinations; identify "on time / delayed / missed dose" based on the opening and closing of the medicine box and the medication plan; and identify "suspected fall" based on acceleration and posture changes. After obtaining the set of elderly care semantic events, the events, states, and contexts are written into a personal elderly care knowledge graph according to the entity and relationship pattern of "time-location-activity-physiology-medication-environment". This forms queryable nodes (such as activities, symptoms, risks) and edges (such as occurrence, impact, association, and sequence constraints), and records the confidence level and evidence index for each node / edge.

[0037] When computer devices are networked with edge devices for edge-side inference, trusted home-based elderly care data and sets of semantic events related to elderly care are injected as evidence into the corresponding nodes / edges of the personal elderly care knowledge graph. Evidence fusion is then performed based on the sequential constraints, causal constraints, or correlation weights in the personal elderly care knowledge graph. Risk reasoning and conclusion generation for elderly care scenarios are then executed. For example, the probability / score and confidence level of target conclusions such as "fall risk, acute abnormalities, decreased medication adherence, abnormal nighttime activity, and prolonged inactivity" are calculated. Each conclusion is then bound to a minimum set of evidence (key time window, key event chain, key state node), ultimately forming a set of conclusions disclosed by the elderly. Each conclusion must include at least the conclusion type, conclusion value, confidence level, evidence index, and interpretable points.

[0038] By comprehensively utilizing the identity credentials, access purpose, and contextual parameters carried in the information access request, and combining them with the current status data of the elderly living at home (e.g., whether they are in a suspected emergency state, whether it is nighttime, or whether the risk has recently increased), a situational situation analysis is conducted to obtain situational risk analysis data and urgency constraints. The basic authorization result (obtained by matching the preset access control policy) and the above situational risk / urgency constraints are input into the situational contract compiler to generate the disclosure level, disclosure scope, validity period, response delay limit, and automatic eviction trigger conditions that apply to this request. The output is information disclosure analysis data, which serves as the sole constraint on "what can be disclosed, to what granularity of disclosure, for how long, and when to eviction."

[0039] Based on the analysis of information disclosure data, the allowed subset of conclusions, field granularity (conclusions / summaries / fragments), time window precision, and frequency limits are determined. Based on this, conclusions are selectively disclosed and arranged to generate home-based elderly care output information. After generating the output information, it undergoes desensitization and trimming to remove content that can reconstruct the original data (e.g., precise trajectories, original sequences, identifiable audio / video clips, etc.), resulting in minimal output information. Based on the minimal output information, conclusion commitment data (proving that the output indeed originates from the disclosed conclusion set), and contextual contract adjudication constraints, verifiable verification materials (such as algorithm / strategy version identifiers, execution environment proofs, timestamps, data commitments / hash codes, and consistency verification results) are generated for the output results. This provides home-based elderly care verification proof, enabling external requesters to verify that the output information was generated at the edge according to contractual constraints and has not been tampered with.

[0040] In the aforementioned IoT-based home-based elderly care information management method, anomaly suppression is performed on the raw home-based elderly care data after receiving information access requests to improve data reliability. Based on trusted data, event semanticization and personal elderly care knowledge graph construction are completed, transforming the home-based elderly care status from "raw collected data" into computable, associative, and interpretable semantic events and structured knowledge. This allows for reasoning at the edge using semantic events and knowledge graphs to obtain a high-confidence set of elderly disclosure conclusions. Furthermore, by introducing contextual contract adjudication oriented towards identity, purpose, and situation, information disclosure analysis data is generated, achieving improved disclosure granularity. Dynamic constraints on scope, timeliness, and retrieval conditions transform the output content from "full data sharing" to "on-demand, minimal" disclosure of conclusions / summaries / fragments. Simultaneously, verifiable minimal disclosure analysis generates home-based elderly care output information and its verification proof, enabling external requesters to verify the source consistency and compliance of the output information without accessing the original reproducible data. This significantly reduces the risk of privacy leaks and unauthorized access, improves the credibility, traceability, and cross-entity collaboration efficiency of home-based elderly care information sharing, reduces cloud or platform-side storage and computing pressure, and enhances the system's continuous service capabilities in weak network or offline scenarios.

[0041] In an exemplary embodiment, based on a set of semantic events related to elderly care and a personal knowledge graph of elderly care, edge-side reasoning is performed on trusted data related to home-based elderly care to obtain a set of conclusions disclosed by the elderly, including steps 302 to 306. Wherein:

[0042] Step 302: Extract the causal skeleton from the set of semantic events related to elderly care and the personal knowledge graph of elderly care to obtain the skeleton diagram of causal constraint relationships.

[0043] Step 304: Extract time-series features from reliable home-based elderly care data to obtain home-based elderly care feature sequence data.

[0044] Step 306: Based on the causal constraint relationship skeleton diagram and the home-based elderly care feature sequence data, counterfactual consistency reasoning is performed on the personal elderly care knowledge graph to obtain the set of conclusions disclosed by the elderly.

[0045] Among them, causal skeleton extraction involves identifying and filtering key causal paths and directional constraints of "event-state-risk" from the set of semantic events related to elderly care and the personal knowledge graph of elderly care, and removing related edges that are only relevant but not causally interpretable.

[0046] Among them, the causal constraint skeleton graph is a sparse graph structure data used to define the acceptable causal structure, which includes nodes (events / states / risks) and their causal direction edges and constraints / weights.

[0047] Among them, time-series feature extraction is the process of aggregating reliable home-based elderly care data by time window and calculating quantifiable indicators such as statistics, trends, rhythms and anomalies to form a time-series feature expression that can be used for reasoning.

[0048] Among them, the home-based elderly care feature sequence data is a feature vector sequence organized by time index, which represents the quantitative results of changes in the elderly’s behavior, physiology, medication and environmental status over time and their reliability weights.

[0049] Counterfactual consistency reasoning involves constructing and comparing inferences from facts and counterfactual scenarios under causal constraints, and then using consistency tests to screen for conclusions that are stable or have significant causal sensitivity.

[0050] Specifically, the event types, sequence relationships, co-occurrence relationships in the semantic event set for elderly care, and the relationship edges (e.g., "cause / induce / correlate / inhibit / occur at") in the personal elderly care knowledge graph are uniformly encoded, and event nodes, state nodes, and risk nodes are constructed into candidate causal graphs. Combining time sequence constraints (e.g., "event A occurs first, then state B appears"), stable covariance constraints (e.g., event chains that recur over multiple days), and existing a priori causal edges in the personal elderly care knowledge graph, candidate causal edges are screened and oriented, removing edges that are only related but do not satisfy sequence or interpretability constraints, and outputting a causal constraint relationship skeleton graph that retains the critical path and constraints of "event-state-risk". Each skeleton edge can be accompanied by direction, strength weight, and applicable conditions (e.g., night / after medication / specific environment).

[0051] The reliable data on home-based elderly care is segmented along a time axis (e.g., by minute / hour / day and key time windows), and feature extraction is performed according to data type. Specifically, for physiological time series, features such as mean, variance, peak value, rate of change, rhythmicity, and duration of abnormalities are extracted. For behavioral data, features such as activity level, number of times getting out of bed, gait stability, sedentary duration, and day / night distribution are extracted. For medication and environmental data, features such as adherence indicators, box opening delay, and trends in temperature / humidity / light / ground humidity are extracted. Simultaneously, features are aligned with event semantic labels (e.g., mapping "frequent nighttime awakenings" to corresponding time window features), ultimately forming a home-based elderly care feature sequence data with time index, feature dimension index, and reliability weight.

[0052] Using a causal constraint framework diagram as the structural constraint and home-based elderly care characteristic sequence data as the evidence input, relevant nodes (events, states, risks) in the individual elderly care knowledge graph are parameterized into inferable variables. Counterfactual scenarios (e.g., "if nighttime light exposure increases / if medication is taken on time / if ground humidity does not increase") are constructed for target conclusions (e.g., increased risk of falls, possibility of acute abnormalities, probability of adverse reactions due to medication). Under the premise of satisfying the framework diagram constraints, the intervention variables are replaced or edges are cut off, and the conclusion probabilities / scores under both factual and counterfactual conditions are calculated. Consistency tests are performed on the conclusions obtained under multiple counterfactual scenarios, and conclusions that remain stable or have significant causal sensitivity under reasonable intervention changes are selected. Each conclusion is then bound to a corresponding key evidence fragment and confidence level, forming a set of conclusions disclosed by the elderly.

[0053] In this embodiment, a causal constraint relationship skeleton diagram is obtained by extracting the causal skeleton from the set of semantic events related to elderly care and the personal elderly care knowledge graph. This transforms subsequent reasoning from "relevance-driven" to "causal constraint-driven," structurally suppressing accidental co-occurrence and noise propagation. By extracting temporal features from credible home-based elderly care data, a home-based elderly care feature sequence data is formed, enabling a quantitative characterization of changes in the elderly's state and alignment with time windows, thus improving the stability and comparability of evidence input. Furthermore, counterfactual consistency reasoning is performed on the personal elderly care knowledge graph under the constraints of the causal skeleton. Conclusions with consistency or significant sensitivity to intervention changes are selected from the comparison of facts and counterfactuals, thereby outputting a set of elderly disclosure conclusions with higher confidence, stronger interpretability, and better robustness, reducing false positives and false negatives, and improving the reliability of decisions regarding disclosure and intervention.

[0054] In an exemplary embodiment, counterfactual consistency reasoning is performed on the personal elderly care knowledge graph based on the causal constraint relationship skeleton diagram and home-based elderly care feature sequence data to obtain the set of conclusions disclosed by the elderly, including steps 402 to 408. Wherein:

[0055] Step 402: Based on the causal constraint skeleton diagram, perform minimum intervention selection on the causal paths in the personal retirement knowledge graph to obtain the counterfactual intervention set.

[0056] Step 404: Extract evidence fragments and perform graph anchoring on key time window data in the home-based elderly care feature sequence data to obtain counterfactual evidence mapping data.

[0057] Step 406: Based on the counterfactual intervention set and counterfactual evidence mapping data, perform probabilistic graphical inference on the personal pension knowledge graph to obtain factual posterior distribution data and counterfactual posterior distribution data.

[0058] Step 408: Perform counterfactual consistency screening on the factual posterior distribution data and the counterfactual posterior distribution data to obtain the set of conclusions disclosed by the elderly.

[0059] Among them, the minimum intervention selection is to select the fewest number of intervenable nodes / edges with the most significant impact on the target conclusion from the causal paths of the personal retirement knowledge graph under the constraint of the causal constraint relationship skeleton graph.

[0060] The counterfactual intervention set is a set of intervention items selected from the minimum intervention. Each intervention item includes at least the intervention object, the intervention value / alternative value (or range), the intervention effective time window, and the constraints.

[0061] Among them, key time window data are data segments located from home-based elderly care characteristic sequence data that correspond to time intervals that are highly correlated with risk changes, event triggers, or sudden changes in state.

[0062] Among them, evidence fragment extraction is the process of extracting evidence subsequences or feature summaries (such as magnitude, duration, rate of change, and confidence) from key time window data that can support reasoning.

[0063] Among them, graph anchoring is the process of matching evidence fragments to corresponding nodes or edges in a personal pension knowledge graph according to entity, relation and time order constraints and establishing an association index.

[0064] Among them, counterfactual evidence mapping data is a structured mapping result that describes the correspondence between "evidence fragment - graph node / edge - time sequence / credibility", which is used to inject observational evidence and constraints during inference.

[0065] Among them, probabilistic graphical inference involves parameterizing the knowledge graph into a probabilistic graph and then combining evidence and intervention to perform inference calculations on the target variable to obtain the posterior distribution (such as belief propagation / variable inference).

[0066] Among them, the factual posterior distribution data is the probability distribution or rating distribution result calculated for the target conclusion variable in a factual scenario without intervention, combined with observational evidence.

[0067] Among them, the counterfactual posterior distribution data is the probability distribution or rating distribution result calculated for the target conclusion variable in the context of counterfactual intervention, combined with observational evidence.

[0068] Among them, the counterfactual consistency screening is a process that compares the factual and counterfactual posterior distributions and filters out unreliable conclusions while retaining disclosable conclusions based on stability / sensitivity and confidence thresholds.

[0069] Specifically, using a causal constraint framework diagram as constraints, the causal paths related to the target conclusion (e.g., increased fall risk, acute abnormalities, decreased adherence) in the individual retirement knowledge graph are traversed and scored. The most sensitive and actionable intervention points (such as nodes or edges like "nighttime light exposure," "medication time deviation," "ground humidity," and "frequency of getting out of bed") are identified without violating the key directional constraints of the framework diagram. Following the "minimum change principle," a set of intervention variables is selected to cover the main causal paths with minimal node / edge changes and maximize conclusion identifiability (e.g., achieving a considerable change in the probability of the conclusion with minimal intervention). A counterfactual intervention set is output, where each intervention item includes at least the intervention target, the range / alternative value of the intervention value, the intervention effective time window, and applicable conditions.

[0070] Key time windows are located in the characteristic sequence data of home-based elderly care. After location, evidence fragments are extracted based on event trigger points, feature mutation points, or risk windows (such as "abnormal activity at night 02:00–04:00", "heart rate fluctuation 30 minutes after medication", "continuous sitting exceeding the threshold"). Summary features (amplitude, duration, rate of change, confidence weight, etc.) are generated for each evidence fragment. The evidence fragments are anchored to corresponding nodes / edges in the personal elderly care knowledge graph through entity alignment and relation matching (for example, anchoring "increased number of times getting out of bed at night" to the "nighttime wake-up event node" and its causal edge with the "fall risk node"). The time sequence constraints, evidence sources, and confidence levels are recorded, ultimately forming counterfactual evidence mapping data.

[0071] Based on the counterfactual intervention set, the personal retirement knowledge graph is subjected to edge severing or conditional probability factor replacement at the intervention nodes, and the intervention values ​​are set as exogenous variables, thus forming the structure and parameters required for counterfactual inference. Simultaneously, the counterfactual evidence mapping data is transformed into probabilistic graphical evidence (e.g., observation factors, soft evidence likelihood, or time window constraint factors) and injected into the corresponding variables of the graph. Then, inference algorithms (such as belief propagation, variational inference, or one or a combination of MCMC) are executed on the "fact graph" (without intervention) and the "counterfact graph" (with intervention), respectively, to calculate the posterior distribution of the target conclusion variable under the observed evidence conditions, outputting the factual posterior distribution data and the counterfactual posterior distribution data, and retaining the evidence contribution or message intermediate value corresponding to each posterior distribution.

[0072] The posterior distribution data of facts and counterfactual posterior distribution data are compared and evaluated. On the one hand, stability indicators of the conclusions under different counterfactual interventions are calculated (e.g., posterior distribution overlap, variance change, consistency of conclusion threshold crossing probability) to screen results that "remain stable to reasonable intervention changes and can be used as reliable disclosure conclusions." On the other hand, causal sensitivity indicators (e.g., ΔP, log-likelihood difference, or change in risk score) are calculated to identify results that "are highly sensitive to specific interventions and can be used to explain causes or provide intervention recommendations." Combining evidence completeness and confidence thresholds, conclusions driven by low-confidence evidence or lacking consistency are eliminated, and the output is a set of disclosure conclusions of the elderly that includes conclusion values, confidence levels, evidence indexes, and corresponding intervention sensitivity / stability labels.

[0073] In this embodiment, by selecting the causal path of the personal elderly care knowledge graph with minimal intervention under the guidance of the causal constraint relationship skeleton graph, key causal links can be covered with a smaller set of intervention variables, thereby reducing the search space of counterfactual reasoning and improving the identifiability of conclusions. By extracting evidence fragments from key time windows of home-based elderly care feature sequence data and anchoring them to the graph, consistent alignment of evidence and graph variables and injection of temporal constraints are achieved, reducing inference bias caused by evidence drift and mismatch. Furthermore, based on the counterfactual intervention set and evidence mapping data, probabilistic graph inference is performed on the knowledge graph to obtain two types of posterior distributions: factual and counterfactual, upgrading risk conclusions from single-point judgments to distributed uncertainty expressions. Finally, through counterfactual consistency screening, conclusions that are stable or have significant causal sensitivity under multiple intervention scenarios are retained, outputting a set of elderly disclosure conclusions with higher confidence, stronger interpretability, and better noise resistance, thereby effectively reducing false positives and false negatives and improving the reliability of disclosure and intervention decisions.

[0074] In an exemplary embodiment, probabilistic graphical inference is performed on an individual's retirement knowledge graph based on a counterfactual intervention set and counterfactual evidence mapping data to obtain factual posterior distribution data and counterfactual posterior distribution data, including steps 502 to 512. Wherein:

[0075] Step 502: Based on the counterfactual intervention set, the personal retirement knowledge graph is intervened and compiled to obtain a counterfactual structure diagram.

[0076] Step 504: Perform evidence factorization processing on the counterfactual evidence mapping data to obtain an evidence factor set.

[0077] Step 506: Based on the evidence factor set, construct the factor graphs corresponding to the personal retirement knowledge graph and the counterfactual structure graph, respectively, to obtain the knowledge graph factor graph and the structure graph factor graph.

[0078] Step 508: Perform belief propagation inference on the knowledge graph factor graph and the structure graph factor graph respectively to obtain the initial fact posterior distribution data and the initial counterfactual posterior distribution data.

[0079] Step 510: Perform coupled message multiplexing processing on the initial fact posterior distribution data and the initial counterfactual posterior distribution data to obtain coupled posterior distribution data.

[0080] Step 512: Based on the causal constraint skeleton diagram, perform causal consistency calibration on the coupled posterior distribution data to obtain the factual posterior distribution data and the counterfactual posterior distribution data.

[0081] Intervention compilation involves applying intervention items from the counterfactual intervention set to the structure and parameters of an individual's retirement knowledge graph, transforming the intervention semantics into an inferable graph structure through edge cutting and / or factor substitution.

[0082] Among them, the counterfactual structure graph is a graph structure representation obtained after intervention and compilation, in which the dependencies and conditional factors of the intervened nodes are rewritten according to the intervention rules, and it is used to describe the causal generation mechanism in the counterfactual scenario.

[0083] Among them, evidence factorization is the process of transforming evidence fragments, credibility and temporal constraints in counterfactual evidence mapping data into observation factors / likelihood factors / constraint factors in probability graphs.

[0084] Among them, the evidence factor set is a set of factors generated by evidence factorization, which is used to represent observational evidence, soft and hard constraints and their weights in a unified form and to attach them to graph variables.

[0085] Among them, the knowledge graph factor graph is a factor graph formed by parameterizing the personal pension knowledge graph. In factual scenarios, it is composed of variable nodes, structural factors, and evidence factors.

[0086] Among them, the structural graph factor graph is a factor graph formed by parameterizing the counterfactual structural graph. In the counterfactual scenario, it includes intervention replacement factors and evidence factors to support counterfactual inference.

[0087] Among them, belief propagation inference is a probabilistic inference method that calculates the posterior distribution or marginal distribution of the target variable through iterative message passing between variables and factors on the factor graph.

[0088] The initial fact posterior distribution data is the fact scenario posterior distribution result obtained by combining evidence factors on the knowledge graph factor graph, and has not yet been coupled, reused, or causally calibrated.

[0089] The initial counterfactual posterior distribution data are counterfactual scenario posterior distribution results inferred from the structural graph factor graph by combining intervention replacement factors and evidence factors, and have not yet undergone coupling reuse and causal calibration.

[0090] Among them, the coupled message reuse process is a process that reuses convergence messages of the uninterrupted region between fact and counterfactual inferences and only recalculates messages of the interfered boundary region to reduce redundant calculations.

[0091] Among them, the coupled posterior distribution data is the posterior distribution result data obtained after coupling message multiplexing, which includes message state information of two types of posteriors, namely fact and counterfactual, and their shared / recalculated boundaries.

[0092] Among them, causal consistency calibration is a process that projects or regularizes the coupled posterior distribution to a feasible set that satisfies the causal direction and path constraints based on the causal constraint skeleton diagram in order to eliminate deviations that violate causal constraints.

[0093] Specifically, based on the intervention target, value / substitute value, and effective time window of each intervention item in the counterfactual intervention set, the corresponding interventioned node or edge in the personal retirement knowledge graph is mapped to an interventionizable variable. Then, a structural-level compilation operation is performed on the interventioned node, that is, the incoming edges representing "cause → effect" are cut off or masked to satisfy the semantics of the do-operation, and the conditional probability factor of the interventioned node is replaced with an exogenous setting factor or a time window condition factor (e.g., fixing "nighttime light exposure" as the target level). For interventions involving time windows, time gating constraints are added to the structure to ensure that the intervention is only effective within the specified interval. Finally, the counterfactual structure graph containing the definitions of the interventioned structural changes and replacement factors is output.

[0094] Each "evidence fragment—graph node / edge—time sequence / credibility" mapping in the counterfactual evidence mapping data is parsed, transforming the evidence fragment into a probabilistic graph-consumable factor representation. Specifically, hard evidence factors (indicator variable values ​​or intervals) are generated for deterministic observations, and soft evidence likelihood factors (e.g., describing the support of the observation to the variable using Gaussian / Bernoulli / piecewise functions) are generated for uncertain observations. The credibility weight of the evidence fragment is encoded as factor strength or temperature coefficient. Temporal consistency factors (e.g., constraint factors for "A occurs first, then B occurs") are generated for time sequence and window constraints, resulting in a set of evidence factors containing observation factors, time sequence factors, and credibility weights.

[0095] The nodes in the personal retirement knowledge graph are instantiated as random variables, and the relationships between nodes / edges (such as causal influence, correlation dependence, and co-occurrence constraints) are mapped as structural factors or conditional probability factors. The evidence factor set is then attached to the corresponding variables according to the graph anchoring results. In factual scenarios, a knowledge graph factor graph is generated, and in counterfactual scenarios, a structural graph factor graph is generated using the structure after intervention and the replacement factors. Consistency checks are performed on the variable domain, factor scope, and time gating of both to ensure that the semantics of the same variable are aligned in the factual and counterfactual factor graphs.

[0096] The prior variables and evidence factors corresponding to the knowledge graph factor graph are used as input initialization messages, and iteratively updated within a "factor-variable-factor" message passing framework. This allows evidence likelihood and structural dependence to gradually propagate to the target conclusion variable. After convergence or reaching the iteration limit, the target variable is marginalized to obtain the initial posterior distribution data of the facts. Simultaneously, intervention replacement factors and evidence factors from the structure graph factor graph are used as input initialization messages, and the process is similar to that of knowledge graph factors. Figure 1 The message scheduling strategy (damping or truncation can be introduced for structures with loops to enhance stability) is used to marginalize the same target variable after iterative update to obtain the initial counterfactual posterior distribution data, and intermediate message buffers and convergence markers are recorded during the inference process.

[0097] Based on the counterfactual intervention set, the variable subgraphs and shared subgraphs affected by the intervention are marked in the knowledge graph factor graph and the structure graph factor graph. Variable-factor messages that have converged within the shared subgraph during the knowledge graph factor graph inference process are injected into the structure graph factor graph as reusable messages, thus skipping duplicate message updates in the shared subgraph in the structure graph factor graph. Then, local message recalculation is performed only on the boundary regions affected by the intervention, and the entire process is completed by concatenating the boundary messages and shared messages. Figure 1 The resulting posterior update outputs coupled posterior distribution data containing both factual and counterfactual inferences, along with corresponding message reuse indexes and boundary recalculation markers.

[0098] The directional constraints, forbidden path constraints, and conditional constraints in the causal constraint skeleton graph are transformed into calibration rules (e.g., restricting certain posterior dependency directions, constraining certain conditional probability ratio ranges, or requiring monotonicity on specific causal paths), and consistency calibration is performed on the coupled posterior distribution data. During consistency calibration, the distributions of variables that violate the skeleton constraints are corrected by constraint projection or regularization (e.g., projecting the distribution to the feasible set using minimum KL divergence), and the messages of boundary variables are re-normalized to eliminate probability leakage introduced by calibration. After calibration, the factual posterior distribution data and counterfactual posterior distribution data that satisfy the causal skeleton constraints are output.

[0099] In this embodiment, a counterfactual structure graph is obtained by intervening and compiling the personal pension knowledge graph based on the counterfactual intervention set. This gives the counterfactual inference clear structured intervention semantics and avoids mistaking correlation for causation. By factorizing the counterfactual evidence mapping data and constructing corresponding factor graphs on the knowledge graph and the counterfactual structure graph, multi-source evidence is injected into the inference process in a unified probability factor form, improving the consistency of evidence fusion. Furthermore, belief propagation is performed on the knowledge graph factor graph and the structure graph factor graph to obtain the initial fact / counterfactual posterior distribution, so that the conclusion expresses uncertainty in a distribution form and improves the interpretability of the inference. On this basis, by reusing the intermediate messages of the uninterrupted area through coupled message reuse, only the intervened boundary area is recalculated, which significantly reduces the redundant calculation overhead of fact and counterfactual dual graph inference and improves the real-time performance of the edge side. Finally, the coupled posterior distribution is calibrated for causal consistency based on the causal constraint skeleton graph to suppress posterior bias that conflicts with the causal skeleton, thereby obtaining fact and counterfactual posterior distribution data that are more consistent with causal constraints, more robust, and more credible.

[0100] In an exemplary embodiment, based on the information access request, a situational contract adjudication is performed on the information disclosure strategy for elderly people living at home to obtain information disclosure analysis data, including steps 602 to 606. Wherein:

[0101] Step 602: Perform situational analysis on the situational parameter data in the information access request and the current status data of the elderly at home to obtain situational risk analysis data and urgency constraint data.

[0102] Step 604: Based on the identity credential data and access purpose data in the information access request, perform permission matching on the preset access control policy corresponding to the elderly at home to obtain basic authorization result data.

[0103] Step 606: Based on the situational risk analysis data, urgency constraint data, and basic authorization result data, perform situational contract compilation and adjudication on the information disclosure strategy to obtain information disclosure analysis data.

[0104] Among them, context parameter data is a set of parameters used in information access requests to describe the environment and interaction context in which the request occurs, including at least the request time, request location / network environment, terminal characteristics, request frequency, and session context.

[0105] The current status data is a collection of real-time or near-real-time home care status data for elderly people living at home, including at least physiological status, behavioral status, medication status, environmental status, and abnormal markers / risk levels.

[0106] Among them, situational analysis is an analysis and processing method that links situational parameter data with current status data to identify risk type, risk level and urgency.

[0107] Among them, situational risk analysis data is the risk outcome data output by situational situation analysis, which represents the risk category, risk score / level, confidence level and key triggering factors.

[0108] Among them, the urgency constraint data is used to constrain the timeliness and urgency rules of disclosure decisions, including at least the urgency level, the upper limit of response delay, the upper limit of disclosure granularity / elevation conditions, and the retrieval trigger threshold.

[0109] Identity credential data refers to authentication materials used to prove the authenticity and credibility of the requester's identity, such as digital certificates, signatures, tokens, device fingerprints, or challenge response results.

[0110] The access purpose data is used to describe the purpose category and task requirements of the requester in accessing home-based elderly care information, such as care, nursing implementation, medical consultation and assessment, or emergency rescue.

[0111] Among them, the preset access control policy is a set of access control rules pre-configured by elderly people living at home or preset by the system, which is used to specify the scope and granularity of allowed access under different identities, purposes and resource types (such as RBAC / ABAC rules).

[0112] Among them, permission matching is the process of mapping identity credential data and access destination data to preset access control policies to perform rule matching and permission calculation.

[0113] The basic authorization result data is the authorization result data output by permission matching, which includes at least the allow / deny flag, accessible resource category, granularity limit, default validity period, and hit rule index.

[0114] Among them, the scenario contract compilation and adjudication is to instantiate and dynamically adjudicate the information disclosure strategy based on the basic authorization result, combined with scenario risk analysis data and urgency constraint data, thereby generating the executable disclosure boundaries and conditions for this request.

[0115] Specifically, the contextual parameter data in the information access request is parsed and vectorized (e.g., request time, geographic / network location, terminal fingerprint, request frequency, session continuity, historical interaction context), and the current status data of the elderly at home is read simultaneously (e.g., recent risk warning level, abnormal vital signs markers, suspected bed-leaning / fall events, medication deviation, prolonged inactivity, environmental hazard factors, etc.). Then, based on a preset risk causal template or rule / model, the "contextual elements - elderly status" are evaluated in a linked manner, outputting contextual risk analysis data (e.g., risk type, risk score, confidence level, key triggering factors) and generating urgency constraint data (e.g., whether it is urgent, upper limit of response delay, upper / lower limit of disclosure granularity, triggering conditions for allowing an upgrade of the disclosure level, and threshold for initiating a retrieval mechanism).

[0116] The system verifies the identity credentials in the information access request (e.g., signature verification, certificate chain verification, device / account binding verification, one-time challenge response verification) and parses the access purpose data to determine the requester's purpose category (care, nursing execution, remote consultation, emergency rescue, etc.) and the required information type. It then invokes the corresponding preset access control policies for elderly people living at home (e.g., role / attribute-based access control ABAC / RBAC policies, elderly individual preference policies, blacklist / whitelist and time period policies) to perform matching calculations on "requester attributes—purpose attributes—resource type / granularity—time range," outputting basic authorization result data, which at least includes allow / deny flags, accessible resource categories, upper limit of accessible granularity, default validity period, and policy hit clause index.

[0117] The scenario risk analysis data, urgency constraint data, and basic authorization result data are input into the scenario contract compiler to instantiate the information disclosure strategy. Within the compiler, abstract clauses in the disclosure strategy (e.g., "only disclose the conclusion if non-urgent," "urgent cases can be temporarily upgraded to a summary and recorded for auditing," "limit location accuracy at night," etc.) are first bound to specific scenario variables of this request, forming an executable set of scenario contract constraints. Then, based on the urgency constraints, the basic authorization result is projected and contracted (e.g., in high-risk emergency scenarios, the disclosure level can be upgraded but the validity period shortened and mandatory recall enforced; in low-risk scenarios, the disclosure granularity reduced and frequency limited), generating a disclosure envelope (disclosure scope, field set, time window precision, frequency cap). Finally, information disclosure analysis data is output, including at least the disclosure level, disclosure scope, disclosure validity period, automatic recall trigger conditions, response latency requirements, and verification requirements.

[0118] In this embodiment, situational analysis data and urgency constraint data are obtained by combining the contextual parameters in the information access request with the current status of the elderly at home. This enables disclosure decisions to perceive dynamic constraints such as "whether it is urgent, how high the risk is, and how fast the response needs to be." Then, based on identity credentials and access purpose, permission matching is performed on the preset access control policy to obtain basic authorization result data, giving the disclosure decision traceable legitimacy and a minimum authorization basis. Finally, under the condition of shrinking / projecting the basic authorization based on situational risk and urgency constraints, the information disclosure strategy is compiled and adjudicated according to the situational contract, generating information disclosure analysis data including disclosure level, scope, validity period, and eviction conditions. This achieves adaptive minimum disclosure and rapid authorization switching according to the situation, significantly reducing the risk of unauthorized access and excessive disclosure, and improving the availability of information in emergency scenarios and the strength of privacy protection in normal scenarios.

[0119] In an exemplary embodiment, situational analysis is performed on the situational parameter data in the information access request and the current status data of the elderly person living at home to obtain situational risk analysis data and urgency constraint data, including steps 702 to 710. Wherein:

[0120] Step 702: Analyze the contextual parameter data in the information access request to obtain contextual element vector data; Step 704: Perform state fragmentation processing on the current state data to obtain a set of state evidence fragments; Step 706: Based on the context element vector data and the set of state evidence fragments, perform matching analysis on the preset risk causal templates corresponding to elderly people living at home to obtain the context risk candidate set and confidence data; Step 708: Perform urgency constraint analysis on the candidate set of situational risks to obtain urgency constraint data; Step 710: Perform risk situation fusion on the candidate set of situational risks and confidence data to obtain situational risk analysis data.

[0121] Among them, context element parsing is the process of extracting and standardizing elements such as time, location / network, terminal fingerprint, request frequency and session context from the context parameters of information access requests.

[0122] Among them, the context element vector data is a vectorized representation formed by numericalizing / discretizing the parsed context elements according to the feature dictionary, which is used for subsequent template matching and scoring calculation.

[0123] Among them, the state fragmentation process is to divide the current state data of elderly people living at home into several fragments with consistent state characteristics according to time windows and change points, and generate fragment summaries.

[0124] The state evidence fragment set is a set of evidence units obtained by state fragmentation processing. Each fragment contains at least a time window, state type label, strength / duration duration, and credibility weight.

[0125] Among them, the preset risk causal template is a set of causal link rules or model templates configured for elderly people living at home, which are "contextual elements - state evidence - risk type / level".

[0126] Among them, matching analysis is a process that checks the constraints and calculates the similarity / score of the context element vector data and the set of state evidence fragments against the preset risk causal template to determine the triggering risk.

[0127] Among them, the scenario risk candidate set is a set of candidate risk items output by the matching analysis, which includes risk type, risk score / level, trigger evidence index and applicable conditions information.

[0128] Among them, confidence data is used to quantify the reliability of each scenario risk candidate, and is usually obtained by combining evidence credibility, template matching degree and consistency test results.

[0129] Among them, the urgency constraint analysis is the process of deriving the urgency level, response delay, disclosure granularity, and recovery conditions based on the type, score, and confidence level of the scenario risk candidate.

[0130] Among them, risk situation fusion is a process of removing redundancy, resolving conflicts, and weighting and aggregating multiple candidate risks and their confidence levels to form a unified and interpretable risk situation output.

[0131] Specifically, the contextual parameter data in the information access request is parsed and standardized. Elements such as request time (e.g., weekday / night / holiday), request location or network location (e.g., home LAN / external network / frequently used locations), terminal and session characteristics (e.g., device fingerprint, IP segment, connection method, session continuity), request frequency and behavior patterns (e.g., short-term repeated requests, abnormally high frequency) are extracted as structured features and then numerically and discretized according to a preset feature dictionary to generate contextual element vector data that can be used for subsequent template matching and scoring. At the same time, credibility or anomaly markers are added to each dimension of the elements to reflect environmental uncertainty.

[0132] The current status data of elderly people living at home is segmented according to the time axis and event trigger points. The continuous status data is divided into several status segments and segment summaries (such as duration, rate of change, peak value, degree of abnormality, and confidence weight) are extracted. The status segments include at least one of the following: physiological abnormality segment (such as persistent abnormal heart rate / blood oxygenation), behavioral abnormality segment (such as frequent getting out of bed at night, long periods of inactivity), medication abnormality segment (such as delayed / missed doses), and environmental abnormality segment (such as insufficient light, increased risk of slippery conditions). At the same time, the segments are deduplicated, merged, and boundary corrected to ensure that the segments are consistent with the actual status changes. Finally, a set of status evidence segments with time windows, type labels, and confidence scores is output.

[0133] Taking contextual element vector data and a set of state evidence fragments as input, the system calls the preset risk causal template library corresponding to elderly people living at home (the templates are represented by a causal link of "contextual element → state fragment → risk type / level" and may include applicable conditions and weights). Then, a matching calculation is performed on each template, that is, first check the hard constraints (such as "night + external network request" and "existence of suspected fall event fragments") and then calculate the soft matching score (such as feature similarity, fragment strength and duration weighted score). The anomaly markers of contextual elements and the credibility weights of fragments are included in the confidence estimation. Finally, the system outputs a set of contextual risk candidates (including candidate risk types, candidate risk levels / scores, and trigger evidence indexes) and the corresponding confidence data.

[0134] An urgency level is generated based on the urgency level rules of candidate risk types in the scenario risk candidate set (e.g., "suspected fall / prolonged lack of response" has the highest priority, while "mild medication delay" has a lower priority), candidate risk scores and confidence thresholds, and time sensitivity (e.g., whether minute-level response is required). At the same time, the constraint parameters that disclosure decisions need to meet are derived, such as the upper limit of response delay, whether temporary increase in disclosure granularity is allowed, the upper / lower limit of disclosure granularity, the minimum necessary fields that can be accessed, and the automatic recovery trigger conditions (e.g., "emergency lifted / time expired / risk score decline"), thereby forming urgency constraint data that can be directly used for scenario contract compilation and adjudication.

[0135] The scenario risk candidate set is clustered and merged according to risk type and time window of action. Duplicate candidates for the same risk under different templates are identified and deduplicated. Then, for conflicting candidate risks (e.g., one candidate points to "low-risk normality" and another to "high-risk emergency"), conflict resolution is performed based on evidence coverage, confidence weight, strength of key state evidence fragments, and anomaly of scenario elements. The dominant risk item is determined, and non-dominant items are marked as "alternative / to be verified". Based on this, the scores of each risk candidate are weighted and aggregated according to confidence data to form a comprehensive risk score. The output includes risk level, main risk type, summary of key evidence chain triggering the risk (including the most critical state evidence fragments and scenario elements), and uncertainty indicators (such as confidence interval or credibility interval), thus obtaining structured and interpretable scenario risk analysis data.

[0136] In this embodiment, situational element vector data is formed by parsing situational parameter data, thereby achieving a structured and quantitative expression of the request environment and interaction context. A set of state evidence fragments is obtained by fragmenting the current state data, allowing key anomalies and changes in the elderly to be extracted into alignable and weighted evidence units. Furthermore, based on the situational element vector data and the set of state evidence fragments, a preset risk causal template corresponding to the elderly at home is matched to obtain a set of situational risk candidates carrying confidence levels. This allows for the identification of risk sources based on "causal links" rather than simple correlations, improving the interpretability of risk identification. Subsequently, urgency constraint analysis is performed on the situational risk candidate set to generate urgency constraint data, ensuring that disclosure decisions have executable constraints such as response timeliness, upper limit of disclosure granularity, and recovery conditions. Finally, situational risk analysis data is output by fusing candidate risks and confidence levels, achieving redundancy removal, conflict resolution, and weighted aggregation of multiple candidate risks, thereby improving the robustness and consistency of situational risk assessment.

[0137] In an exemplary embodiment, based on situational risk analysis data, urgency constraint data, and basic authorization result data, a situational contract compilation and adjudication is performed on the information disclosure strategy to obtain information disclosure analysis data, including steps 802 to 810. Wherein:

[0138] Step 802: Based on the situational risk analysis data and urgency constraint data, compile the situational contract for the information disclosure strategy to obtain the situational contract constraint set data; Step 804: Solve the disclosure envelope of the context contract constraint set data to obtain the disclosure envelope data; Step 806: Based on the disclosed envelope data, perform permission projection matching on the basic authorization result data to obtain the projected authorization result data; Step 808: Based on the urgency constraint data, construct the timeliness recovery conditions corresponding to the projection authorization result data to obtain the disclosure validity period data and automatic recovery trigger condition data; Step 810: Based on the disclosure validity period data and the automatic recycling trigger condition data, the situational contract constraint set data is instantiated to obtain information disclosure analysis data.

[0139] Among them, the contextual contract compilation is a process that combines the abstract rules in the information disclosure strategy with contextual risk analysis data and urgency constraint data to bind parameters and rewrite conditions to generate executable constraints.

[0140] Among them, the context contract constraint set data is the output data of the context contract compilation, which is used to describe in the form of rules / constraints the objects, granularity, scope, frequency, precision and proof requirements that are allowed to be disclosed in the current context.

[0141] Among them, the disclosure envelope solution is a process of synthesizing and optimizing multiple constraints in the context contract constraint set to obtain the maximum set of disclosable boundaries that satisfy the constraints.

[0142] Among them, the disclosure envelope data is the boundary description data obtained by solving the disclosure envelope, which represents the types of information that can be disclosed, the field whitelist, the time window range and precision, the frequency limit and the list of prohibited items.

[0143] Among them, permission projection matching is a process that maps and shrinks the basic authorization result to the disclosable boundary defined by the disclosure envelope, and obtains the final executable authorization by taking the intersection.

[0144] Among them, the projection authorization result data is the authorization result data output by permission projection matching, which includes the resource objects that are finally allowed to be accessed, the disclosure granularity, the field set, the time window precision, the frequency limit, and the hit rule index.

[0145] Among them, the time-limited recovery conditions are a set of constraints used to control the authorization lifecycle, which stipulate the length of the authorization validity period and the monitoring conditions and actions that trigger revocation / downgrade / scope reduction.

[0146] Among them, the disclosure validity period data is parameter data describing the length of time or expiration time of the projection authorization result during the operation period, which can be set in layers according to resource sensitivity and disclosure granularity.

[0147] Among them, the automatic recycling trigger condition data describes the conditional rules data that automatically execute recycling / downgrading when specific trigger signals are met (such as risk removal, session end, abnormal frequency, credential expiration, withdrawal of consent, etc.).

[0148] Among them, the adjudication instantiation is the process of solidifying the situational contract constraint set, disclosure validity period, and automatic recovery trigger conditions into a disclosure decision instance that can be directly executed in this request and outputting information disclosure analysis data.

[0149] Specifically, using scenario risk analysis data and urgency constraint data as contextual input, abstract clauses in the information disclosure strategy (such as "only conclusions are disclosed under normal circumstances," "statistical summaries can be disclosed for high-risk situations," and "temporary granularity increases are allowed in emergencies, but auditing and recovery are mandatory") are parameter-bound and conditionally rewritten. During the binding and rewriting process, scenario variables such as risk type / risk level, triggering evidence category, urgency level, and response delay limit are first replaced in the strategy condition expression. Then, resource objects in the strategy (conclusions / summaries / fragments, field sets, time window precision, frequency limits) are mapped to computable constraints. Simultaneously, conflicting clauses are merged according to priority (urgency priority, seniority preference priority, minimum disclosure priority, etc.), ultimately outputting an executable set of scenario contract constraints. Each constraint includes at least the applicable conditions, allowed disclosure targets, granularity boundaries, frequency / precision limits, and proof requirements.

[0150] The intersection / union relationships of multiple constraints in the contextual contract constraint set data are transformed into a boundary description of a "disclosure space". This involves constraining and synthesizing multiple restrictions (such as fields, time windows, precision, and frequency) on the same resource object to obtain the maximum disclosure set that simultaneously satisfies all hard constraints in the current context. Soft constraints are then optimized using a weighted priority or minimum disclosure objective. The solution is output in envelope form, including at least the set of allowed information types, the set of allowed fields, the allowed disclosure time window range and precision, the upper limit of allowed disclosure frequency, and a list of prohibited disclosure items, thus obtaining the disclosure envelope data.

[0151] Using the disclosure envelope data as the "disclosure upper bound," the allowed resource categories and granularity limits in the basic authorization result data are mapped to the fields / time windows / precision / frequency dimensions defined in the disclosure envelope. An intersection operation is performed to obtain the executable permissions when both "legitimate authorization" and "contextual constraints" are met. When the allowed range of the basic authorization is greater than the disclosure envelope, shrinkage is performed (reducing granularity / reducing fields / reducing time windows / reducing frequency). When the basic authorization is stricter, the basic authorization restrictions are not relaxed, and the projected authorization result data is output, which at least includes the resource objects that are finally allowed to be accessed, the disclosure granularity, the set of fields, the time window precision and frequency limits, and the corresponding strategy hit index to support auditing.

[0152] The urgency level, response delay limit, and risk persistence parameters in the urgency constraint data are mapped to the disclosure validity period limit. Combined with the resource sensitivity and disclosure granularity in the projection authorization result data, the validity period is stratified, assigning shorter validity periods to sensitive resources (such as location and explanatory fragments) and longer validity periods to low-sensitivity resources (such as statistical summaries), thus obtaining disclosure validity period data. The trigger signals defined in the urgency constraint data (such as risk score decline, emergency event resolution, session end, abnormal request frequency, identity credential expiration, and elderly person withdrawing consent) are compiled into monitorable Boolean conditions or threshold rules. Each type of trigger signal is bound to a corresponding retrieval action (such as downgrading disclosure granularity, narrowing disclosure scope, or directly revoking authorization) and priority, thus obtaining automatic retrieval trigger condition data. Finally, a correlation index is established between the disclosure validity period data, the automatic retrieval trigger condition data, and the projection authorization result data, enabling real-time monitoring and automatic execution of authorization retrieval or downgrading based on expiration time and trigger conditions during operation.

[0153] Using disclosure validity period data and automatic recycling trigger condition data as "runtime constraint injection items," the scenario contract constraint set is solidified from an abstract set of rules into a disclosure decision instance that can be directly executed in this information access request. Specifically, constraint clauses consistent with the projected authorization result data are selected from the scenario contract constraint set data, and parameters such as the allowed disclosure objects, disclosure granularity, field whitelist, time window range / precision, and frequency limit are written into the instantiation result. Next, the disclosure validity period data is written into the instance's validity period field, and the automatic recycling trigger condition data is written into the instance's recycling condition field. Simultaneously, corresponding recycling actions (revocation / downgrade / scope reduction) and trigger priorities are bound to each recycling condition. Finally, a unified disclosure level identifier and disclosure scope description (for subsequent minimum disclosure selection and desensitization trimming) are generated for the instantiation result, and proof requirements and audit tag indexes are attached, thus outputting information disclosure analysis data containing "disclosure content boundaries + effective validity period + recycling mechanism + proof requirements."

[0154] In this embodiment, by incorporating situational risk analysis data and urgency constraint data into the situational contract compilation of the information disclosure strategy, an executable situational contract constraint set is formed. This allows disclosure rules to be dynamically parameterized according to the level of risk and urgency, rather than being statically fixed. Further, the constraint set is used to solve for the disclosure envelope, thereby unifying the constraint fields, granularity, time window precision, and frequency with a "disclosureable space boundary" to avoid over-disclosure. Then, the disclosure envelope is projected onto the basic authorization result to obtain the projected authorization result, realizing the intersection of the two types of constraints: "authorization legality" and "minimum situational disclosure." Subsequently, based on the urgency constraint, time-limited recovery conditions are constructed for the projected authorization, generating disclosure validity periods and automatic recovery trigger conditions. This enables disclosure permissions to have a lifecycle control that allows for revocation, downgrading, and automatic recovery. Finally, the validity period and recovery conditions are injected into the constraint set to complete the adjudication instantiation and output information disclosure analysis data. This strengthens privacy protection in normal scenarios, ensures information availability in emergency scenarios, and improves the executability, traceability, and anti-unauthorization capabilities of disclosure decisions.

[0155] In an exemplary embodiment, based on information disclosure analysis data, a verifiable minimum disclosure analysis is performed on the set of disclosure conclusions for the elderly to obtain home-based elderly care output information and home-based elderly care verification proof, including steps 902 to 910. Wherein:

[0156] Step 902: Based on the information disclosure analysis data, perform disclosure budget analysis on the set of disclosure conclusions for the elderly to obtain the minimum disclosure budget data and proof template data; Step 904: Based on the minimum disclosure budget data, perform conclusion commitment binding on the elderly disclosure conclusion set to obtain conclusion commitment data; Step 906: Based on the proof template data and conclusion commitment data, selectively disclose and arrange the set of conclusions disclosed by the elderly to obtain home-based elderly care output information; Step 908: Desensitize and trim the output information of home-based elderly care to obtain the minimized output information; Step 910: Verify the minimized output information based on the minimized output information, the conclusion commitment data, and the information disclosure analysis data to obtain the home-based elderly care verification certificate.

[0157] Among them, disclosure budget analysis is the process of deconstructing the disclosure level, scope, accuracy, frequency and proof requirements into computable constraints based on information disclosure analysis data, and determining the minimum necessary output boundary for this disclosure.

[0158] Among them, the minimum disclosure budget data is the set of budget parameters output by the disclosure budget parsing, which represents the minimum disclosure configuration such as the types of conclusions that can be disclosed, field whitelists, time window range and precision, frequency limit and prohibited items.

[0159] Among them, the proof template data is the proof specification data that discloses the budget parsing output, which is used to define the types of verification elements to be generated, input and output fields, proof generation rules and verification interfaces.

[0160] Among them, the conclusion commitment binding is a process that calculates the commitment value for the subset of conclusions that are allowed to be disclosed based on the minimum disclosure budget and binds it to the adjudication information to form a verifiable and non-replaceable reference.

[0161] Among them, the conclusion commitment data is the commitment result data generated by the conclusion commitment binding, which includes the commitment identifier (such as hash / Merkelgen) of the conclusion subset and its verifiable reference path and associated metadata.

[0162] Selective disclosure orchestration is a process that, under the constraints of proof templates and conclusion commitments, selects and assembles allowed disclosure fields from the set of elderly disclosure conclusions to generate provable and verifiable structured output content.

[0163] Among them, desensitization and cropping are processes that delete, generalize, quantify, or replace fields in the output information that may lead to re-identification or restoration of the original data, in order to meet the requirements of minimum disclosure and non-restoreability.

[0164] The minimized output information is the version of the external output information obtained after desensitization and trimming, which only contains the conclusions / summaries / fragments that are allowed to be disclosed and does not contain the content that can restore the original home-based elderly care data.

[0165] Specifically, the disclosure level, disclosure scope (field whitelist / time window precision / frequency limit), validity period, and proof requirements are extracted from the information disclosure analysis data and deconstructed into computable disclosure constraint tuples (such as "conclusion only / summary allowed", "only the most recent 24 hours", "location accuracy ≥ street level", "no more than N times per hour", etc.). Then, the set of disclosure conclusions for the elderly is divided into disclosable semantic slices according to conclusion type and interpretable elements, and privacy risks and information gains are assessed for each semantic slice to determine the disclosure cost. Under the premise of satisfying disclosure constraints, the output field set, time window set, precision level, and frequency configuration are solved with the goal of minimizing disclosure, forming the minimum disclosure budget data. Simultaneously, according to the proof requirements (e.g., consistency proof, contract constraint proof, timeliness proof, algorithm / strategy version identifier), the corresponding proof components are selected and the proof input / output interface is instantiated to generate proof template data.

[0166] Based on the minimum disclosure budget data, the subset of conclusions allowed for disclosure and their corresponding evidence index range are determined. A commitment value is calculated for each conclusion within this subset (e.g., hashing / Merkelizing "Conclusion Identifier + Conclusion Value + Time Window + Evidence Index + Version Identifier"). All commitment values ​​are then organized into a verifiable data structure (such as a Merkel root or a commitment list). Simultaneously, the contract instance identifier, validity period, and disclosure scope summary from the information disclosure analysis data are bound to the commitments to ensure a strong correlation between the commitments and "this ruling," ultimately yielding the conclusion commitment data.

[0167] Based on the proof template data, the required field set, field granularity, and proof element type and generation rules for each field are determined. Then, based on the conclusion commitment data, the commitment identifier and verifiable citation path for allowed conclusion disclosure are determined. According to the conclusion type, field whitelist, time window range, and accuracy requirements limited by the minimum disclosure budget data, the conclusion items in the elderly disclosure conclusion set are filtered, retaining only allowed conclusion values, confidence indices, and allowed explanatory elements. The filtered content is assembled according to a preset output format, forming a structured record for each conclusion. Each output field in the record embeds a proof element placeholder matching the proof template data and binds commitment citation information consistent with the conclusion commitment data. This ensures that the output field can generate proof according to the proof template data in subsequent verification stages and be located and validated through the conclusion commitment data, thus obtaining home-based elderly care output information that meets the disclosure budget constraints.

[0168] Using the "granularity upper limit, precision upper limit, and prohibited items list" in the minimum disclosure budget data as the tailoring criteria, the structure of the home-based elderly care output information is processed field by field through "reduction-generalization-quantification-replacement". For example, fields exceeding the allowed granularity are directly deleted (such as original sequences, precise trajectories, and identifiable audio and video features); fields that can indirectly restore privacy are generalized or quantified (such as rounding precise timestamps to the allowed time granularity, generalizing coordinates to blocks / communities, and changing continuous values ​​to intervals or levels); sensitive entities in explanatory content are masked (such as addresses, contact persons, and specific door numbers); and "summary replacement" is used for evidence descriptions, replacing descriptions that can reconstruct the original data with event tags, statistical indicators, or risk factor contribution levels. After tailoring, the output field set is again checked for consistency with the disclosure budget whitelist to ensure that the output does not contain content that can restore the original data and does not exceed the allowed disclosure range, thus obtaining the minimum output information.

[0169] The consistency check is calculated for each output field in the minimized output information using the conclusion commitment data (e.g., verifying Merkel path, commitment hash matching, and field not exceeding boundaries) to ensure that the output does indeed come from the committed subset of conclusions. Secondly, the contract constraints such as disclosure level, scope, validity period, and frequency are verified against the information disclosure analysis data (e.g., verifying that the output field set is consistent with the whitelist, the time window precision is not refined, and the validity period has not expired). The above consistency results are then encapsulated and signed / authenticated according to the proof template along with the contract instance identifier, timestamp, and algorithm / strategy version identifier, ultimately obtaining a home-based elderly care verification certificate that can be verified by external requesters.

[0170] In this embodiment, the disclosure budget of the elderly disclosure conclusion set is analyzed based on information disclosure analysis data to obtain minimum disclosure budget data and proof template data. This ensures that the disclosure content has a calculable "minimum necessary" boundary in terms of field, granularity, time window precision, and frequency, and simultaneously determines the generation specifications of verifiable proof. Furthermore, the conclusions are committed to the minimum disclosure budget to form conclusion commitment data, establishing an irreplaceable and traceable link between subsequent external outputs and the original conclusion set, thus suppressing tampering and splicing at the source. Then, the conclusion set is selectively disclosed and arranged using the proof template and conclusion commitment to generate output information, achieving structured delivery of "output by budget + verifiable by template." Subsequently, the output information is anonymized and trimmed to obtain minimal output information, significantly reducing the privacy risks of re-identifying and restoring the original data. Finally, the minimal output information, conclusion commitment data, and information disclosure analysis data are used for verification and to generate verification proof. This allows external requesters to verify the consistency of the source of the output information, contract compliance, and completeness without accessing the original home-based elderly care data, thereby improving the credibility, traceability, and cross-entity collaboration efficiency of the disclosure results.

[0171] In an exemplary embodiment, based on information disclosure analysis data, a disclosure budget parsing is performed on the elderly disclosure conclusion set to obtain minimum disclosure budget data and proof template data, including steps 1002 to 1010. Wherein:

[0172] Step 1002: Deconstruct the disclosure level data, disclosure scope data, and verification requirement data in the information disclosure analysis data to obtain a set of disclosure constraint tuples; Step 1004: Based on the set of disclosure constraint tuples, perform disclosable semantic slice analysis on the conclusion type space of the elderly's disclosure conclusion set to obtain a semantic slice set. Step 1006: Perform a privacy risk quantification assessment on the semantic slice set to obtain privacy consumption weight data; Step 1008: Based on the set of disclosure constraint tuples and privacy consumption weight data, perform budget optimization on the semantic slice set to obtain the minimum disclosure budget data; Step 1010: Based on the set of disclosure constraint tuples and the minimum disclosure budget data, the verification proof requirement data is templated to obtain the proof template data.

[0173] The disclosure level data is used to indicate the form and granularity of information that is allowed to be disclosed to the public, such as only conclusions, statistical summaries, or explanatory fragments.

[0174] The disclosure scope data is used to limit the boundaries of the content that can be disclosed this time, and includes at least a field whitelist, time window range and precision, frequency limit and prohibited items list.

[0175] Among them, the verification proof requirement data is a set of data used to specify what kind of verification materials need to be generated and their mandatory metadata and verification rules, such as consistency proof, contract compliance proof, and timeliness proof.

[0176] Among them, constraint deconstruction is the process of breaking down disclosure levels, disclosure scopes, and verification requirements from abstract strategies into computable object-relationship-threshold / range constraints.

[0177] The disclosure constraint tuple set is a set of constraints obtained by constraint destructuring, where each tuple describes an executable disclosure restriction and its applicable conditions in a structured form.

[0178] Among them, the conclusion type space is the classification and attribute domain consisting of all conclusion categories and their interpretable element dimensions in the set of conclusions disclosed by the elderly.

[0179] Among them, the disclosable semantic slice analysis is an analysis process that, under the constraint of the set of disclosure constraint tuples, splits the conclusion type space into the smallest disclosure semantic units that can be independently selected and pruned.

[0180] The semantic slice set is a set of candidate units obtained from disclosable semantic slice analysis. Each slice corresponds to a set of output fields, allowed time window precision, and explanatory element combinations.

[0181] Among them, the privacy risk quantitative assessment is an assessment process that measures the sensitivity, linkability, and inferability of each semantic slice and calculates the privacy cost.

[0182] Among them, the privacy consumption weight data is a weight parameter output from the privacy risk quantitative assessment, which is used to represent the relative privacy consumption caused by disclosing each semantic slice.

[0183] Among them, the budget optimization solution is a solution process that selects the optimal output combination from the semantic slice set with the goal of minimizing privacy consumption, under the premise of satisfying disclosure constraints and purpose coverage requirements.

[0184] Template instantiation is the process of solidifying the verification proof requirements, the set of disclosure constraint tuples, and the minimum disclosure budget into a proof template structure and rules that can be directly generated and verified.

[0185] Specifically, when performing constraint deconstruction on the disclosure level data, disclosure scope data, and verification requirement data in the information disclosure analysis data, the process involves three steps: First, mapping the disclosure level to the allowed output format (conclusion / statistical summary / explanation fragment) and default granularity boundaries. Second, breaking down the disclosure scope into specific restrictions such as field whitelists, time window ranges and precision, frequency limits, and prohibited item lists. Third, breaking down the verification requirements into the required proof type, mandatory metadata items (such as contract instance identifier, validity period, version identifier, and timestamp), and proof verification interface parameters. Finally, these restrictions are uniformly represented as a computable set of disclosure constraint tuples, where each tuple contains at least the constraint object, constraint relationship, threshold / range, and applicable conditions.

[0186] A semantic dimension coordinate system is constructed by categorizing the set of disclosure conclusions for the elderly into conclusion types (e.g., fall risk, medication adherence, activity / sleep indicators, environmental risks, etc.) and interpretable element dimensions (e.g., trigger event labels, key time window summaries, evidence indexes, confidence level expressions). Then, the fields and granularity constraints in the disclosure constraint tuple set are projected onto this coordinate system to obtain a set of semantic slices that can be selected as the smallest disclosure unit under the current constraints. Each semantic slice corresponds to a set of output fields, allowed time window precision, and a set of optional interpretable elements.

[0187] For each semantic slice in the semantic slice set, its privacy consumption weight is calculated. This involves weighting each semantic slice's field sensitivity (e.g., location, precise time, continuous physiological sequence, identifiable event chain), linkability (the possibility of re-identification after splicing with external information), inferability (the degree to which details of the original data can be deduced from the slice), and sample sparsity (smaller samples are easier to identify). This weighting is dynamically adjusted in conjunction with the elderly's individual privacy preferences and the cumulative amount of historical disclosures. The result is a privacy consumption weight data that corresponds one-to-one with each semantic slice, representing the privacy cost of disclosing that slice.

[0188] The inviolable hard constraints in the disclosure constraint tuple set are transformed into feasibility conditions to limit the optional states of each semantic slice (e.g., whether output is allowed, the subset of fields allowed for output, and the allowed time window precision and frequency limits). Privacy consumption weight data is assigned as a cost function to each optional state. With "minimum coverage required to meet the request purpose" as the constraint objective (e.g., must cover a specified conclusion type or must include certain required fields), a combination is selected from feasible semantic slices and their optional states to minimize the total privacy cost while satisfying the hard constraints and coverage constraints. If multiple solutions with the same cost exist, the combination with coarser disclosure granularity, wider time window, or fewer fields is preferred. Finally, the selected semantic slice combination and its corresponding field whitelist, time window range / precision, frequency configuration, and prohibited item verification rules are solidified and output as the minimum disclosure budget data.

[0189] The template extracts the contract constraints that must be covered by the proof (such as disclosure level, scope, validity period, and summary of redemption conditions) and metadata items related to the proof environment (such as contract instance identifier, timestamp, strategy / algorithm version identifier, and requester identifier summary) from the set of disclosure constraint tuples, and defines them as the "mandatory fields" of the template. Then, the whitelist of fields, time window precision, frequency cap, and prohibited item rules in the minimum disclosure budget data are compiled into "rule verification expressions" within the template, used to directly determine whether the output exceeds the limits during verification. Simultaneously, a corresponding proof structure is configured for each type of output, including at least "output field consistency verification with whitelist," "output consistency verification with conclusion commitment data (commitment ID / path placeholder)," and "time validity verification," and the input positions and output formats are defined for these verification items. Finally, proof template data that can be directly filled by the proof generation module or executed directly by an external verifier is generated.

[0190] In this embodiment, a set of disclosure constraint tuples is formed by deconstructing the disclosure level, disclosure scope, and verification requirements, transforming the disclosure rules from abstract descriptions into computable and executable constraint expressions. Furthermore, under the constraints of the disclosure constraint tuple set, a set of semantic slices is obtained by performing disclosable semantic slicing analysis on the conclusion type space of the elderly's disclosure conclusion set, thus breaking down complex conclusions into controllable minimum disclosure units, facilitating refined tailoring. Next, a privacy risk quantification assessment is performed on the semantic slice set to obtain privacy consumption weight data, making the privacy cost of different disclosure units measurable and comparable. Subsequently, budget optimization is performed by combining disclosure constraints and privacy weights to obtain minimum disclosure budget data, minimizing privacy consumption and suppressing over-disclosure while satisfying contractual hard constraints and purpose coverage requirements. Finally, based on the disclosure constraints and minimum disclosure budget, template instantiation is performed on the verification requirements to generate proof template data, enabling subsequent outputs to be verified in a unified manner for field compliance, commitment consistency, and timeliness validity, thereby simultaneously improving the executability, auditability, and verifiability of minimum disclosure.

[0191] In one exemplary embodiment, a computer device is provided, which may be a computer device, and its internal structure diagram may be as follows: Figure 3 As shown. This computer device includes a processor, memory, input / output interfaces (I / O), and communication interfaces.

[0192] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0193] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0194] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0195] In one embodiment, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and executes the computer instructions, causing the computer device to perform the steps in the above method embodiments.

[0196] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0197] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0198] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for managing home-based elderly care information based on the Internet of Things, characterized in that, The method includes: In response to an information access request for an elderly person living at home, obtain the original home-based elderly care data corresponding to that elderly person. Data anomaly suppression processing is performed on the original home-based elderly care data to obtain reliable home-based elderly care data; Based on the reliable data on home-based elderly care, the home-based elderly care status is analyzed using event semantics to obtain a set of elderly care semantic events and a personal elderly care knowledge graph. Based on the set of semantic events related to elderly care and the personal knowledge graph of elderly care, edge-side reasoning is performed on the trusted data of home-based elderly care to obtain a set of conclusions disclosed by the elderly. Based on the information access request, a situational contract adjudication is performed on the information disclosure strategy of the elderly living at home to obtain information disclosure analysis data. Based on the information disclosure analysis data, a verifiable minimum disclosure analysis is performed on the set of elderly disclosure conclusions to obtain home-based elderly care output information and home-based elderly care verification certificates.

2. The method according to claim 1, characterized in that, The step involves performing edge-side reasoning on the trusted home-based elderly care data based on the set of elderly care semantic events and the personal elderly care knowledge graph to obtain a set of conclusions disclosed by the elderly, including: The causal skeleton is extracted from the set of semantic events related to elderly care and the personal knowledge graph of elderly care to obtain a skeleton diagram of causal constraint relationships. Temporal features are extracted from the trusted home-based elderly care data to obtain home-based elderly care feature sequence data; Based on the causal constraint skeleton diagram and the home-based elderly care feature sequence data, counterfactual consistency reasoning is performed on the personal elderly care knowledge graph to obtain the set of conclusions disclosed by the elderly.

3. The method according to claim 2, characterized in that, The step involves performing counterfactual consistency reasoning on the personal elderly care knowledge graph based on the causal constraint skeleton diagram and the home-based elderly care feature sequence data to obtain the set of conclusions disclosed by the elderly, including: Based on the causal constraint skeleton diagram, the minimum intervention selection is performed on the causal paths in the personal retirement knowledge graph to obtain a counterfactual intervention set. Evidence fragments are extracted and graph anchoring is performed on key time window data in the home-based elderly care feature sequence data to obtain counterfactual evidence mapping data. Based on the counterfactual intervention set and the counterfactual evidence mapping data, probabilistic graphical inference is performed on the personal pension knowledge graph to obtain factual posterior distribution data and counterfactual posterior distribution data. The counterfactual consistency screening is performed on the factual posterior distribution data and the counterfactual posterior distribution data to obtain the set of conclusions disclosed by the elderly.

4. The method according to claim 3, characterized in that, The step of performing probabilistic graphical inference on the individual retirement knowledge graph based on the counterfactual intervention set and the counterfactual evidence mapping data to obtain factual posterior distribution data and counterfactual posterior distribution data includes: Based on the counterfactual intervention set, the personal retirement knowledge graph is intervened and compiled to obtain a counterfactual structure diagram; The counterfactual evidence mapping data is subjected to evidence factorization processing to obtain an evidence factor set; Based on the set of evidence factors, construct the factor graphs corresponding to the personal retirement knowledge graph and the counterfactual structure graph, respectively, to obtain the knowledge graph factor graph and the structure graph factor graph. Belief propagation inference is performed on the knowledge graph factor graph and the structure graph factor graph respectively to obtain the initial fact posterior distribution data and the initial counterfactual posterior distribution data. The initial fact posterior distribution data and the initial counterfactual posterior distribution data are subjected to coupled message multiplexing processing to obtain coupled posterior distribution data. Based on the causal constraint skeleton diagram, the coupled posterior distribution data are calibrated for causal consistency to obtain the factual posterior distribution data and the counterfactual posterior distribution data.

5. The method according to claim 1, characterized in that, The step of adjudicating the information disclosure strategy of the elderly person living at home based on the information access request, and obtaining information disclosure analysis data, includes: Contextual situation analysis is performed on the contextual parameter data in the information access request and the current status data of the elderly person living at home to obtain contextual risk analysis data and urgency constraint data; Based on the identity credential data and access purpose data in the information access request, the preset access control policy corresponding to the elderly person at home is matched for permissions to obtain basic authorization result data. Based on the scenario risk analysis data, the urgency constraint data, and the basic authorization result data, the scenario contract compilation and adjudication of the information disclosure strategy are performed to obtain the information disclosure analysis data.

6. The method according to claim 5, characterized in that, The situational analysis of the contextual parameter data in the information access request and the current status data of the elderly person living at home yields situational risk analysis data and urgency constraint data, including: Context element parsing is performed on the context parameter data in the information access request to obtain context element vector data; The current state data is processed into state fragments to obtain a set of state evidence fragments; Based on the situation element vector data and the set of state evidence fragments, a matching analysis is performed on the preset risk causal template corresponding to the elderly living at home to obtain a set of situation risk candidates and confidence data. An urgency constraint analysis is performed on the candidate set of scenario risks to obtain the urgency constraint data; The scenario risk candidate set and the confidence data are fused to obtain the scenario risk analysis data.

7. The method according to claim 5, characterized in that, The step of performing a situational contract compilation and adjudication on the information disclosure strategy based on the situational risk analysis data, the urgency constraint data, and the basic authorization result data to obtain the information disclosure analysis data includes: Based on the scenario risk analysis data and the urgency constraint data, the information disclosure strategy is compiled into a scenario contract to obtain scenario contract constraint set data. The disclosure envelope is solved by performing disclosure envelope calculation on the scenario contract constraint set data to obtain disclosure envelope data; Based on the disclosed envelope data, permission projection matching is performed on the basic authorization result data to obtain the projected authorization result data; Based on the urgency constraint data, construct the timeliness recovery conditions corresponding to the projection authorization result data to obtain disclosure validity period data and automatic recovery trigger condition data; Based on the disclosure validity period data and the automatic recycling trigger condition data, the scenario contract constraint set data is instantiated to obtain the information disclosure analysis data.

8. The method according to claim 1, characterized in that, The step involves performing a verifiable minimum disclosure analysis on the set of disclosed conclusions about the elderly based on the information disclosure analysis data, to obtain home-based elderly care output information and home-based elderly care verification proof, including: Based on the information disclosure analysis data, the disclosure budget is analyzed on the set of disclosure conclusions for the elderly to obtain the minimum disclosure budget data and proof template data. Based on the minimum disclosure budget data, conclusion commitment binding is performed on the elderly disclosure conclusion set to obtain conclusion commitment data; Based on the proof template data and the conclusion commitment data, the set of conclusions disclosed by the elderly is selectively disclosed and arranged to obtain the home-based elderly care output information. The output information for home-based elderly care is anonymized and trimmed to obtain the minimized output information; The minimized output information, the conclusion commitment data, and the information disclosure analysis data are used to verify the minimized output information and obtain the home-based elderly care verification certificate.

9. The method according to claim 8, characterized in that, The step involves analyzing the disclosure budget of the elderly disclosure conclusion set based on the information disclosure analysis data to obtain minimum disclosure budget data and proof template data, including: The disclosure level data, disclosure scope data, and verification requirement data in the information disclosure analysis data are constrained and destructured to obtain a set of disclosure constraint tuples. Based on the set of disclosure constraint tuples, a disclosable semantic slice analysis is performed on the conclusion type space of the elderly disclosure conclusion set to obtain a semantic slice set. A privacy risk quantification assessment is performed on the semantic slice set to obtain privacy consumption weight data; Based on the set of disclosure constraint tuples and the privacy consumption weight data, the semantic slice set is optimized to obtain the minimum disclosure budget data. Based on the set of disclosure constraint tuples and the minimum disclosure budget data, the verification proof requirement data is templated to obtain the proof template data.

10. A home-based elderly care information management system based on the Internet of Things, characterized in that, The system includes: computer equipment and triggering terminal; The triggering terminal is used to respond to an information access request for an elderly person living at home and obtain the original home-based elderly care data corresponding to the elderly person; the original home-based elderly care data is transmitted to the computer device via a network; The computer equipment is used to perform data anomaly suppression processing on the original home-based elderly care data to obtain reliable home-based elderly care data. The computer device is used to perform event semantic analysis on the home-based elderly care status of the elderly based on the trusted home-based elderly care data, and obtain a set of elderly care semantic events and a personal elderly care knowledge graph. The computer device is used to perform edge-side reasoning on the trusted home-based elderly care data based on the set of elderly care semantic events and the personal elderly care knowledge graph, to obtain a set of conclusions disclosed by the elderly. The computer device is used to make a situational contract ruling on the information disclosure strategy of the elderly living at home based on the information access request, and obtain information disclosure analysis data. The computer device is used to perform verifiable minimum disclosure analysis on the set of elderly disclosure conclusions based on the information disclosure analysis data, and to obtain home-based elderly care output information and home-based elderly care verification certificates.

Images